310 results about "Row vector" patented technology

The invention discloses a processor and a method for executing matrix multiplication on the processor. The processor comprises a data bus and an array processor composed of k processing units. The data bus is used for reading n columns of row vectors in sequence from M*N multiplicand matrixes, inputting the row vectors to the processing units in the array processor, reading n*k sub-matrixes from N*K multiplier matrixes, inputting column vectors of the sub-matrixes into the corresponding processing units in the array processor, and outputting results obtained through multiplication executed by the processing units. Each processing unit in the array processor is used for executing parallel vector multiplication on the input row vectors and the column vectors, and each processing unit comprises a Wallace tree multiplying unit composed of n multiplying units and n-1 adders. The implementation mode improves processing efficiency of matrix multiplication.

The invention discloses a binary chop tracking method of shortest paths in a network map, which is characterized in that the method comprises the following steps: extracting the step number of the shortest paths from a source node to a gathered node according to a shortest path step number matrix; reducing and separating the step number by half into the front step number and the back step number; searching nodes with the step number equal to the front step number in a source node row vector of the shortest path step number matrix, using the nodes as a front step number node set, searching nodes with the step number equal to the back step number in a gathered node column vector of the shortest path step number matrix, using the nodes as a back step number node set; taking an intersection set of the front step number node set and the back step number node set as middle nodes; using each node in the intersection as the reference; using the method for solving the middle node of the source node and the node, and the middle node of the node and the gathered node until the shortest path step number between the solved nodes is 1; and sequentially inserting the middle nodes into corresponding positions to obtain all shortest paths from the source node to the gathered node. The method of the invention is used for finding all shortest paths between two points, and the invention can provide various schemes for the preferential path selection of the network map abstracted from actual data.

An encoding method by which an encoding speed is improved is disclosed. An encoder (100) comprises an input data storage section (107) for outputting the stored input data (D100) according to an output control signal (108), an input data count section (101) for counting the inputs of input data (D100), an output control section (102) for controlling the output destination of the input data (D100)according to the input counts, one-bit storage sections (103-1 to 103-(N-K)) for holding one-bit data, row-vector storage sections (104-1 to 104-K) for holding row vectors of an LDPC code creation matrix, vector multiplying sections (105-1 to 105-K) for multiplying a row vector and a column vector, a parity storage section (109) for holding a parity created by the multiplication, and an LDPC code-word series creating section (106) for creating an LDPC code word from the input data series and parity series and outputting it.

Disclosed is a V-BLAST system for a MIMO communication system. In the V-BLAST system for a MIMO communication system, a pseudo inverse matrix calculator receives a channel transfer function matrix including channel information and produces a cofactor matrix and a determinant for a pseudo inverse matrix. A norm & minimum calculator calculates a minimum index for the cofactor matrix outputted from the pseudo inverse matrix calculator, a weight vector selector selects a row vector having the minimum index and calculates a transposed matrix for the row vector; an adder adds the transposed matrix to a received input symbol, and a subtractor subtracts the determinant to the output. A demapper performs a determined function operation to the output and produces estimated information.

The invention provides a method for interleaving external code coding output codons, which comprises the steps of: obtaining a first matrix according to a codon array of the external code coding output codons, wherein the line number of the first matrix is the number of the external code coding output codons and the row number is the codon length of the external code coding output codons; dividing the first matrix into at least four sub matrixes; and carrying out line-row vector transposition combination on at least two sub matrixes in the at least four sub matrixes to obtain input codons of internal code coding. The invention also provides a method for deinterleaving external code coding output codons and an interleaving device and a deinterleaving device. By adopting the methods for interleaving and deinterleaving external code coding output codons and the interleaving and the deinterleaving devices, external code coding and internal code coding are subjected to parallel interleaving cascade, and therefore, random errors and / or longer outburst errors of the internal code coding are uniformly distributed in the external code coding output codons, and an error flat bed of cascade codes is lowered.

The invention provides an ISAR (inverse synthetic aperture radar) imaging orientation calibration method, which comprises the following steps that a radar records echo to obtain echo data using the distance as the line vector and the orientation as the row vector; distance units are selected for forming a new echo matrix; a frequency domain windowing technology is adopted for extracting the scattering point echo; the phase quadratic term factor of the distance unit echo is estimated; the orientation resolution ratio of the ISAR image is calculated on the basis of a least squares estimation method; the orientation calibration on the ISAR images is completed according to the orientation resolution ratio of the ISAR image. The method provided by the invention has the advantages that 2-order DPT conversion is adopted for estimating the phase quadratic term factor of the distance-dimension echo, the defects of real-time application inconvenience and great calculation quantity during the phase quadratic term estimation of the traditional method based on LPFT and image contrast ratio are overcome, and in addition, the rotating angular speed of a target is obtained through the least squares estimation method, so the orientation calibration of the ISAR images is realized.

The invention relates to a two-dimensional direction of arrival estimating method based on a planar array and an estimator. The two-dimensional direction of arrival estimating method includes the steps that a decoupling model ( F=AL (alpha)*G (beta)+N ) of planar array received signals is built based on the first direction of arrival alpha and the second direction of arrival beta, wherein F represents the planar array received signals, AL (alpha) represents a manifold matrix related to the first direction of arrival alpha, G (beta) represents a signal matrix related to the second direction of arrival beta, and N represents a noise signal matrix; on the premise of alpha domain sparsity, alpha domain sparse reconstruction is carried out on the planar array received signals, so that a support set of alpha domain sparse signals and an alpha estimated value are obtained; on the premise of beta domain sparsity, beta domain sparse reconstruction is carried out on a row vector, corresponding to the alpha estimated value, in the signal matrix G (beta), so that a support set of beta domain sparse signals and a beta estimated value corresponding to the alpha estimated value are obtained. Alpha and beta are sequentially estimated by using the spatial domain sparsity, in an alpha domain and a beta domain, of the incident signals, and therefore computation complexity is reduced substantially.

The invention provides an automatic extraction method and device for stable characteristics of a power system. The method includes: determining sample input quantity and classification number of a logistic regression training model and defining a superparameter set; defining cost function containing LASSO penalty terms and updating a logistic regression parameter matrix theta in the cost function with the LASSO penalty terms by the aid of an iteration method; selecting the logistic regression training model generated by the maximum superparameter with the error rate smaller than a threshold in the superparameter set as the optimum model; taking the maximum value of sum of the row vector absolute value of the logistic regression parameter matrix theta as the stable characteristics of a power grid. The critical clearing time (CCT) of the stabilization rate of the logistic regression parameter model on expected failure of the power grid is matched, the penalty items of 1 norm of the parameter matrix theta are added in the cost function, and the main characteristics are highlighted; the stable characteristics are automatically extracted.

The present invention provides a detection method of an unstructured point cloud feature point and an extraction method thereof. The extraction method includes (1) calculating the Harris response value of a sampling point in different scale space; (2) selecting the Harris response value of the optimal scale space as the Harris response value of the sampling point to obtain a feature point set Q; (3) selecting one maximum point of the Harris response values possessing maximality in both of the scale space neighborhood and a geometric neighborhood as a candidate feature point, at last, selecting the optimizing strategy to draw the final feature point. A tangent plane of the gained feature point is subjected to network segmentation under a polar coordinate system, and then a neighborhood point of the feature point is projected to the tangent plane, a feature information statistical matrix is generated by voting projected length corresponding to projective points from each grid to four peaks of the grid, then both of a row vector and a column vector are respectively subjected to the DCT transform and the DFT transform, the elements of the upper left corner after transform is a character description vector.

A channel state information (CSI) reporting method is provided. A reference signal is received from a base station. A precoding matrix is selected, based on the reference signal, from a codebook, where the codebook includes a precoding matrix W and W=αSV, where a matrix V is an N×ν matrix, N is the number of reference signal ports, ν≦N, S is a row selection matrix used to select one or more row vectors from the matrix V, and α is a constant. CSI is sent to the base station, where the CSI includes a precoding matrix indicator PMI and the PMI is corresponding to the selected precoding matrix. A suitable precoding matrix can be selected according to an interference situation, so as to select an antenna for transmitting data and power of the antenna.

The invention discloses a ground-based radar multi-target detection method based on compressed sensing. The method includes the steps that sparse representations of echo signals are completed, and a sparse transform matrix is schemed out; a random Gauss matrix is chosen as an observation matrix, and each row vector in the observation matrix is used for projecting target echo signals; part of signal information is obtained, and the echo signals undergo reduced-order observation; the sparse representations of the target echo signals are set as optimization variables, and a convex optimization model is established and solved to obtain a global optimum solution of the sparse representation signals; according to the sparse matrix construction method and the sparse representations of the echo signals, a target time-lag and a target Doppler frequency are obtained. Compared with a traditional orthogonal matching method, the method can correctly detect a large number of targets when the number of targets is uncertain.

The invention discloses a copy video detection method based on a continuous copy frame sequence, which arms at providing a method of effectively detecting a splicing condition of copy and non copy fragments of a to-be-detected video. In the technical scheme, key frames and key frame features of a reference video and a to-be-detected video are extracted firstly to obtain a reference video key framefeature matrix and a to-be-detected video key frame feature matrix; row vectors with the minimum Euclidean distances are searched in the reference video key frame feature matrix for each row in the to-be-detected video key frame feature matrix to obtain the most similar reference video key frame sequence number list R<*> and the most similar reference distances d1, d2,...,di,...,dM; and based onthe R<*> and d1, d2,...,di,...,dM, whether the to-be-detected video key frame is a copy frame is judged, whether the to-be-detected video is a copy video is judged, and the copy fragment in the copy video is positioned. By adopting the method in the invention, the condition in which the to-be-detected video is spliced by a copy fragment and a non-copy fragment can be effectively solved, and the similar key frame matching speed and the accuracy are improved.

A method for generating a codeword that is insensitive to variations of a channel and easily extensible is provided. The method includes obtaining a primary unit by taking a row vector from an N(E(N−1) simplex code and mapping a control signal to the codeword. The simplex code is obtained by removing a first row from an N(ENorthogonal matrix. The codeword is obtained by combining a plurality of primary units.

The invention discloses a failure mode recognition method of an epicyclic gearbox using a mixed domain feature vector and grey correlation analysis. The method comprises the steps that (1) the feature parameters ft, ff, fs and ftf of the time domain, the frequency domain, the order domain and the time-frequency domain are extracted to constitute the mixed domain feature vector Fm, and the weighted values of wt, wf, ws and wtf of all the feature parameters are calculated; (2) the vibration monitoring health state of the epicyclic gearbox and the historical data of various failure mode states are acquired, and the corresponding mixed domain feature vector value mj(k) is calculated to be used as a reference state matrix; (3) the feature vector t of a signal to be detected and the row vector of the reference state matrix m are both used as the input of a grey correlation analysis algorithm; (4) the correlation vector of the feature vector t of the signal to be detected and the standard mode matrix m is calculated to be used as an output vector of the grey correlation analysis algorithm. The failure mode recognition method of the epicyclic gearbox using the mixed domain feature vector and grey correlation analysis has the advantages of being simple in principle, simple and convenient to operate, stable and reliable, high in precision, suitable for high interference and varied working conditions and the like.

The invention discloses a method for generating long scrambling code phase shift code, and a relative method, wherein the method comprises that using long scrambling code phase shift to find the phase rotary matrix of long scrambling code sequences X, Y, according to the matrix in the step A, rotating vector to obtain phase shift generation vector, according to the phase shift generation vector in the step B, rotating the long scrambling code sequences X, Y to obtain the sequence value after phase shift, generating the sequence values after phase shift, to generate long scrambling code with shift phase. The invention only uses four row vectors to complete the generation of long scrambling code phase shift code. Compared with register shift and matrix rotation long scrambling code sequence phase methods, the invention uses less data, with high speed, simple system, and easy application.

The invention relate to a wireless sensor network key pre-distribution method, which comprises the following steps: step S001: disposition server generating R upper triangular matrixes G and R diagonal matrixes D at random on the finite field GF (q); step S002: generating R groups of real symmetric matrixes K; step S003: saving the ID of every node, row vector seed gir, column vector; step S004: starting base station, and clustering networking; step S005: discovering the shared secret key pair corresponding to secret key space between nodes by pre-distributing secret key; step S006: distributing cluster secret keys with cluster head towards inner-cluster by the constructed shared secret key pair between nodes; step S007: expanding the secret key space dimension. The invention achieves lower node capacity, communication complexity and calculation complexity while ensuring the network completely is communicated; and has higher node anti-capture capability.

A parity check matrix making it possible to encode through decoding, a method of generating a parity check matrix, an encoding method and an error correction apparatus including defining an M×N parity check matrix H=[Hm|Hp], and generating an M×M matrix as a sub-matrix Hp wherein all row vectors are linearly independent, a set A of all of the row vectors is a union set of non-empty subsets A1, A2, . . . , Ak (1≦k≦M) that do not include intersection sets with each other, A1 is a set of weight one row vectors, and Ai (2≦i≦k) is a set of row vectors capable of deriving a weight one row vector by a linear combination with row vectors in a union set of the subsets A1, . . . , Ai-1 among the row vectors not included in the union set.

The invention discloses a CNN text classification method combined with a multi-head self-attention mechanism, and the method comprises the steps: obtaining a word segmentation sequence of a to-be-classified text, and carrying out the filtering of special symbols and stop words; querying a preset or randomly initialized word embedding model to obtain an embedding matrix of the word segmentation sequence, wherein each line of the matrix is an embedding vector of each segmented word; for each row vector of the obtained embedded matrix, superposing the position coding vector of the word segmentation corresponding to the vector; generating a self-attention matrix for the embedded matrix after superposition position coding through a self-attention mechanism; repeating the generation of the self-attention matrix for a plurality of times, and splicing the plurality of generated matrixes on the dimension of the column; multiplying the self-attention matrix after word sequence splicing by the weighting matrix to realize dimensionality reduction and fusion; and inputting the dimensionality-reduced and fused self-attention matrix into a CNN for training or prediction.

The invention relates to a system and a method for video compression sensing and encoding based on measurement of domain block classification. The system comprises a block compression sampling module, a block classification module of a measurement domain, a sampling rate allocation module, a sampling number allocation and secondary measurement module and a reconstruction module. An encoding end divides an input video block into sub-blocks which are equal in size and not mutually overlapped, and each of the sub-blocks is measured by using the same random measurement matrix to obtain measurement domain signals; the block classification module of the measurement domain uses the measurement domain signals as row vectors of the matrix to obtain a measurement domain characteristic image, a correlation relation model between the measurement domain signals and frequency domain signals is built, judgment of the block classification is carried out on the measurement domain characteristic image according to the relation model, sampling rates are allocated to different types of video blocks, the sampling number allocation and secondary measurement module determines a measurement number and a measurement matrix according to the sampling rates and the video blocks, video signals are measured again by using the measurement matrix to obtain measurement signals, and the measurement signals are transmitted to the reconstruction module to reconstruct.

A method and an apparatus for recognizing an intention, and a non-transitory computer-readable recording medium are provided. The method includes learning vectors of knowledge base elements in corpus samples, and converting the corpus samples into row vectors composed of the vectors of the knowledge base elements in a knowledge base; extracting feature vectors from respective pooling windows in the corpus samples by hierarchical pooling, determining weights positively correlated with similarities between texts within the respective pooling windows and the respective corpus samples, performing weighting on the extracted feature vectors to obtain feature vectors of the respective pooling windows, and obtaining feature vectors of the respective corpus samples composed of the feature vectors of the pooling windows; training a vector-based intention recognition classifier, based on the feature vectors of the corpus samples; and recognizing an intention in querying a corpus, using the trained intention recognition classifier.

The invention proposes a hybrid beamforming method based on row vector optimization. The algorithm comprises the following steps: step a, establishing a hybrid pre-coding problem model of a millimeterwave MIMO system; step b, calculating an optimal pre-coding matrix by a transmitting end according to channel information and randomly generating an analog pre-coding matrix; step c, solving a digital pre-coding matrix by the transmitting end under an alternately optimized frame; step d, solving the analog pre-coding matrix by the transmitting end under the alternately optimized frame by using amanifold optimization method; and step e, outputting digital pre-coding and analog pre-coding of a transceiving end. The method proposed by the invention can be used for solving the hybrid pre-codingproblem in the millimeter wave MIMO system under multiple situations. The proposed algorithm can achieve higher spectral efficiency with lower complexity, and has certain innovation and practicability.

Circuitry can be provided in a fixed logic device, or can be configured into a programmable integrated circuit device such as a programmable logic device (PLD) to perform vector-based matrix operations for matrices of arbitrary size, up to a predetermined maximum size. The circuitry may be used where input row vectors of a matrix are to be combined—e.g., by multiplication—with the same initial vector, which may be one of the rows. Column memories may be provided to hold the input matrix data, so that each row may be read by simultaneously accessing the same index in each column memory. In accordance with the invention, the number of column memories may be less than the actual number of columns so that multiple physical “row access” operations are used to access each logical row. A “circular latch” may be provided to hold the initial vector.

The invention provides a data predicting model selecting device and method. The model selecting device comprises a matrix decomposition unit and a model selecting unit, wherein the matrix decomposition unit is configured to perform singular value decomposition on a prediction matrix, the row / column vectors of the prediction matrix are sample vectors, the column / row vectors of the prediction matrix are feature vectors of the corresponding dimensions, and one of the sample vectors is a vector to be predicted, which contains a feature sequence to be predicted; the model selecting unit is configured to select a model to be applied to prediction of the feature sequence to be predicted on the basis of the submatrix of the sample vector, which is obtained through singular value decomposition and can reflect the prediction matrix. The data predicting model selecting device and method can reduce matrix computation dimensionality, reduce noise and achieve the effects of automatically and accurately selecting the prediction model.

The invention discloses a constructive method of a multi-dimensional orthogonal paseudorandom matrix, which is applied to spread spectrum of a CDMA system. The steps of the method are as follows: primitive polynomials, f1, f2, L, fn are screened and combined according to requirements, and fn representation sequences after being screened and assembled are constructed into orthogonal paseudorandom matrix M through group translating in a buffer memory; threshold values are set, then through row comb filters and column comb filters, mutual orthogonal excellent code blocks are selected out from the matrix as permissible codes, and remaining forbidden codes are filtered. The multi-dimensional orthogonal paseudorandom matrix of the invention is characterized by simple method and circuit structure. Sequence sets in the matrix have good orthogonal property, paseudorandom property and ultra-large scale property. The multi-dimensional orthogonal paseudorandom matrix M is applied in the CDMA, which uses column vectors as channel coding; code division multiple access is applied in the CDMA, and row vectors are used as user codes. Simultaneously, the invention can also be used in Ad Hoc, and the matrix is used for dividing base stations and the column vectors and the row vectors in the matrix are used as the channel coding and user codes in every base station simultaneously.

The invention relates to a cipherkey setting method of a heterogeneous wireless sensor network. The method comprises the following steps: step1, three layers of heterogeneous network structures are built; step2, a plurality of t-order symmetric binary polynomials are generated on a galois field GF (q), and S t-order symmetric binary polynomials form an N*N symmetric matrix A; each strong-capacity node stores a certain row vector j of an upper triangular matrix and a column vector j of a corresponding lower triangular matrix, meanwhile, the row number j of the certain row vector j is used as an ID number to be stored, and a Hash function is stored; step3, the cipherkeys of the strong-capacity nodes (H-sensor) and common sensor nodes (L-sensor) are pre-distributed; step4, the session cipherkeys among the strong-capacity nodes (H-sensor) are generated; and step5, the session cipherkeys of the strong-capacity nodes (H-sensor) and the common sensor nodes (L-sensor) are generated. Compared with the prior art, the method has the advantages that the relationship between the energy consumption and the safety of the wireless sensor network are further balanced; and on the basis that the storage consumption of the whole network is reduced, the safety of the network is increased, and the network nodes realize a 100-percent ability to resist capture attack.

The invention discloses an ammeter court identification method. The method includes: selecting two transformers in each day of the test period; respectively connecting high-power electric equipment toa power supply line to which the transformer belongs, respectively starting one piece of electric equipment in different test intervals in the test time period and enabling the electric equipment tocontinuously work until all transformers in the transformer set at least complete one round of test, and acquiring voltage data of all to-be-identified ammeters to obtain an acquired data matrix; summating every 3Y elements of each row vector in the acquired data matrix to obtain a summation matrix; performing binary transformation calculation based on the summation matrix to obtain a binary matrix with the same dimension; combining the single-round test binary matrixes obtained in the test period; and identifying the court affiliation of each to-be-identified ammeter based on the similarity between the row vectors of the binary matrix and the merging matrix. According to the method, the transformer area attribution of the electric meter can be analyzed by collecting the test voltage dataof the electric meter, and the method has the advantages of low engineering cost, convenience in operation and high automation level.

The invention provides a vehicle fuel consumption evaluation method and device. The method comprises the steps of acquiring N pieces of driving behavior data to determine an evaluation object set; acquiring the evaluation weight value and the membership evaluation vector corresponding to each piece of driving behavior data in the N pieces of driving behavior data; constructing evaluation weight vectors based on the evaluation weight values corresponding to the N pieces of driving behavior data acquired by an acquiring module, performing normalization processing on the evaluation weight vectors, and constructing a membership evaluation matrix by using the membership evaluation vector corresponding to each piece of driving behavior data as a row vector; acquiring an evaluation result vector according to an evaluation weight vector obtained after normalization processing and a membership evaluation matrix obtained after normalization processing, and determining the current fuel consumption level of a vehicle according to the evaluation result vector. By collecting, analyzing and calculating driving behavior data, the fuel economy, namely the fuel consumption level, of a vehicle can be graded; the method and the device can solve the problem of excessive fuel consumption caused by bad driving behavior.

The invention discloses a transformer on-load tap-changer mechanical condition monitoring method. The method includes collecting a vibration signal during the switching of a transformer on-load tap-changer; calculating the embedding dimension of the vibration signal; calculating the scale factor of the vibration signal, and performing coarse graining processing on the vibration signal so that a reconstruction matrix can be obtained; performing dimensionality reduction and decomposition on the reconstruction matrix; performing symbolization and coding processing on the row vector in the reconstruction matrix after dimensionality reduction so that the shape characteristic sequence of the vibration signal can be obtained; and calculating the similarity of the shape characteristic sequence ofthe vibration signal of a to-be-detected tap-changer and the shape characteristic sequence of a history vibration signal, and judging the mechanical condition of the tap-changer through the similarity. The method can effectively and highly sensitively monitor the mechanical conditions of transformer on-load tap-changers on line, and therefore, the timely overhaul or replacing of the tap-changers can be realized, and the operation reliability of the tap-changers and transformers can be enhanced.

The invention relates to a resource allocation prediction method and system for mobile cloud computing. The method comprises the steps of performing statistics on resource index values in mobile cloudcomputing in a preset time period before a target time node to form an index value set A, and performing grading on the resource index values in the set A, wherein each grade corresponds to a state space; by utilizing a weighted Markov chain algorithm, calculating a state probability row vector of the target time node; according to grading standards, establishing mean column vectors of the resource index values, and in combination with the state probability row vector, calculating a predicted value of the target time node. According to the method, the predicted value can be directly obtained,so that a model is simplified to a certain extent, the calculation amount is reduced, and the efficiency is improved. By utilizing the mean column vector, the influence of an interval superposition probability on a whole prediction result is reduced, the prediction accuracy is improved, and the problem that the final prediction result deviates from an actual result due to the fact that the weighted Markov chain algorithm ignores difference of prediction probabilities in stages in the superposition process is solved.

The invention provides a multi-dimensional control method and device based on gestures. The control method comprises detecting a gesture action of a user so as to obtain an electromyographic signal Zcorresponding to the gesture action, wherein the gesture action is at least a linear combination of a first single-degree-of-freedom action and a second single-degree-of-freedom action; and processingthe obtained electromyographic signal Z according to the following equation: Z'=W'.F, wherein Z' is a characteristic signal of the electromyographic signal Z, and W 'is a pseudo-inverse matrix of a system transfer matrix W, and F is a non-negative control signal matrix, the system transfer matrix is obtained through a sparse non-negative integer factor decomposition algorithm in the training process, and the row vector of the non-negative control signal matrix and the corresponding single-degree-of-freedom action are synchronous and are in direct proportion. The multi-dimensional control method and control device based on the gestures can convert the electromyographic signal representing the gesture actions into a control signal based on sparse non-negative matrix factor decomposition, soas to carry out synchronization and direct proportion control based on the gestures.